Machine learning predicts cartilage proteoglycan content from MR fingerprinting

Preliminary study: In bovine cartilage, Gaussian process regression predicted proteoglycan content from MR fingerprinting (median r=0.81, NRMSE 11.7%), but collagen fiber anisotropy prediction was weak (r=0.40). Raw MRF data outperformed qMRI maps.

• Proteoglycan content predicted with median r=0.81 and NRMSE 11.7%; raw MRF data outperformed quantitative MRI maps.
• Collagen fiber anisotropy prediction was weak (median r=0.40, NRMSE 26.4%) with no difference between raw MRF and relaxation time maps.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.